Researchers used AI to design a new material that they used to build a working battery – it requires up to 70 percent less lithium than some competing designs.
They designed and built a battery that uses up to 70 per cent less lithium than some competing designs.
This is probably a way of phrasing that means it’s up to 70% less than the absolute most lithium-requiring designs that few/no one uses, and probably only marginally better than most designs actually used. Since they’re very vague about it, I will be sceptical and assume it is way less revolutionary than the headline suggests.
Also, lithium is of pretty low concern when it comes to the materials in current cells. Stuff like cobalt and nickel are more critical and would be larger news.
LFP batteries are both nickel and cobalt free, and are being used in production cars right now (e.g. Tesla model 3/Y standard range options). That technology has long arrived.
Yes, also Lithium Manganese Spinel cells have been around since 1996 and also don’t contain any nickel and cobalt. This is good but many vehicles and devices still use NMC and NCA due to the better specific energy density which is where LFP is limited (but can output more power and is much safer). Tesla (and every EV manufacturer) compromises on the battery depending on what chemistry they use, where if they could reduce the need for expensive metals while maintaining specific energy it would be pretty newsworthy.
this work does nothing to address this, and they also include yttrium, because they focus on solid electrolytes for some reason (probably because chemical space is smaller)
Also, AI would have just sped up an existing plan they had to try new approaches because AI doesn’t create new ideas or think of things out of nowhere.
If you tell AI to do things within a certain range and it gives you results then AI came up with a design as much as google came up with search results when you put something into the search bar.
It can apply existing concepts in ways we haven’t thought of. AI has been used for exactly this thing for years in chemistry. When given constraints (less lithium) and parameters (with this much capacity) it can try permutations of various designs that theoretically meet those conditions.
Yes AI is overhyped, yes it’s often exaggerated by news sources, but that doesn’t mean AI is a non-invention or something. It’s a long way off from any of the lofty goals that are often thrown around by tech ceos, but that doesn’t mean it’s useless.
It can apply existing concepts in ways we haven’t thought of, like people do. AI has been used for exactly this thing for decades in chemistry. When given constraints (less lithium) and parameters (with this much capacity) it can try permutations of various designs that theoretically meet those conditions.
We have had weather models, astronomical models, and all other kinds of computer based prediction methods that do multiple permutations that theoretically meet conditions. AI is just another step forward by doing better pattern recognition and identifying relationships with data based on design choices. All of the chemistry findings came from the system being designed to try things they would not normally test for because testing is expensive and AI can run simulated tests faster and cheaper.
My point is that saying ‘AI came up with’ is 100% inaccurate phrasing intended to trick people into thinking that AI is intelligent instead of just being a very complex tool used to do things we already do faster. It allows for trying more permutations and more pattern recognition, but is just another approach to existing computer models that have also identified things we did not expect. Computer models used to identify starts with planets, but we don’t call those intelligent because they aren’t being sold as something they are not.
Ah, I see what you’re saying. Yes the recognition for these advances should be with human programmers and engineers who are configuring the software and making the models for testing. You’re right I can definitely see why that distinction is important and the media should be making clear that the AI isn’t just turned on and magically works it all out on its own. It’s computational resources being directed towards a task, the models it works within are setup by professionals and the discoveries it finds are interpreted and made useful by those professionals.
The media is just parroting what the companies that want to sell AI are saying. They suck at reporting anything technical or scientific for sure, but they didn’t come up with this on their own.
Your first comment my first thought was how does this have any upvotes. Thats super wrong
Top notch comback with this comment, i still cant agree with the original wording, i do recognize your point and agree with yoiur sentiment. Its a tool first and foremost.
That’s not true at all. AI can in fact generate novel techniques and solutions and has already done so in biotech and electrical engineering. I don’t think you understand how AI works or what it is
I think maybe people are running into a misunderstanding between LLMs and neural nets or machine kearning in general? AI has become too big of an umbrella term. We’ve been using NNs for a while now to produce entirely new ways to go about things. They can find bugs in games that humans can’t, been used to design new wind turbine blades (even made several asymmetrical ones which humans just don’t really do), or plot out entirely new ways of locomotion when given physical bodies. Machine learning is fascinating and can produce very unique results partly because it can be set up to not have existing design biases like humans do
And the nature of computers is that they are magnitudes better than humans at brute forcing. Machine learning can brute force (depending on the technique, it can be smarter than brute forcing, being more efficient) test many many many more designs and techniques than we could manually do. Sure it’ll fail many times, but it’s just a numbers game, and it can pump those numbers. It’ll try a lot of weird and unique stuff we wouldn’t even think to try, with varying degrees of success.
Name one that wasn’t just doing the thing it was told and the users being surprised. You know, the same way that people are surprised when research has results they did not expect using other approaches.
It’s a weird way of asking this. Of course it’s going to do what’s told, the alternative is that it, out of the blue, spits a battery design for no reason. If it were to somehow find a way to make batteries with less lithium in a way that never did before, isn’t that an unexpected result using other approaches?
This is not general artificial intelligence, everything we have is narrow AI, focused on solving one specific problem, for identifying birds to understand instructions between drugs.
That’s the point, it takes all the factors we know about and speed runs through all the possible ways it could work. Humans don’t have the time to look for every single possible way a battery could be constructed, but a ML model can just work it’s way through the issue faster and without human intervention.
Plus just like with the new group of antibiotics we just used AI to discover, it will allow truly thinking Humans to expand upon it.
Really sick of this “oh but you don’t realize AI don’t actually think! Therefore it’s all worthless!” With this smug bullshit like you think you’re bringing anything of value to the conversation.
I didn’t say it was worthless. In fact, I said the exact same things you just said in another post but with the additional detail that the name actually does matter when it is clearly misleading people into thinking it is something that it is not.
What a terribly ignorant thing to say, when people make these armchair comments they’re only hurting ordinary people that can make real benefits from using the technology.
What a giant leap you have taken there. Speeding up existing processes is an extremely helpful thing for the average people, just like weather models that also did things we were already doing far faster and with more variables than people could handle without the automation.
AI will be very helpful. It will not magically solve all of our problems on its own, which is how ‘AI comes up with’ is being presented.
By this very same logic, nobody has ever discovered anything because they’re just speeding someone else’s plans of improving or deriving from someone else’s findings
At the core, weather models, web searches, and AI are all pattern recognition with various levels of complexity and scope. Just like a bicycle is comparable to a motorcycle because they both have two wheels even though one is powered and can go faster and for longer without wearing out the rider.
By this very same logic, nobody has ever discovered anything because they’re just speeding someone else’s plans of improving or deriving from someone else’s findings
AI is not a person capable of coming up with something on its own.
I never claimed that, humans discoveries are just new permutations of observed phenomena. Every single mechanism of the universe is a permutation of the baseline functionality: physics. Therefore, if we’re shitting on permutations, you’re shitting on all of science. AI can do what we do faster. It’s just applied “knowledge” - no different than humans. In fact, that’s the whole point of neural networks, to emulate what our brains literally do right now.
you would know that if you read the article. they replaced part of lithium in electrolyte with sodium, so that they can use less lithium. the problem is decreased ion mobility ie less power density in real life terms.
Baker and Murugesan both say that lots of work is left to optimise the new battery.
bet
i’m gonna mostly ignore this finding because it sounds like extension of AI hype. real lab work is still absolutely critical in order to make it work
And then there’s a hundred other factors. How many charge cycles does it get? Cold weather performance? Can it be mass produced? Does it improve safety over current cells?
It might be useful for what it leads to. Batteries get better because we explore ten different options and then one of them works out. People have gotten less excited over individual discoveries like this for mostly fair reasons. But then there’s another layer of understanding beyond that where you see it as one path of many.
This is probably a way of phrasing that means it’s up to 70% less than the absolute most lithium-requiring designs that few/no one uses, and probably only marginally better than most designs actually used. Since they’re very vague about it, I will be sceptical and assume it is way less revolutionary than the headline suggests.
Also, lithium is of pretty low concern when it comes to the materials in current cells. Stuff like cobalt and nickel are more critical and would be larger news.
LFP batteries are both nickel and cobalt free, and are being used in production cars right now (e.g. Tesla model 3/Y standard range options). That technology has long arrived.
Yes, also Lithium Manganese Spinel cells have been around since 1996 and also don’t contain any nickel and cobalt. This is good but many vehicles and devices still use NMC and NCA due to the better specific energy density which is where LFP is limited (but can output more power and is much safer). Tesla (and every EV manufacturer) compromises on the battery depending on what chemistry they use, where if they could reduce the need for expensive metals while maintaining specific energy it would be pretty newsworthy.
Yeah, for cars, energy density is the name of the game. We honestly don’t need more output power and Tesla is not one to care about safety lol.
But indeed for grid storage, those chemistries are much more useful where energy density is less critical.
this work does nothing to address this, and they also include yttrium, because they focus on solid electrolytes for some reason (probably because chemical space is smaller)
Also, AI would have just sped up an existing plan they had to try new approaches because AI doesn’t create new ideas or think of things out of nowhere.
If you tell AI to do things within a certain range and it gives you results then AI came up with a design as much as google came up with search results when you put something into the search bar.
It can apply existing concepts in ways we haven’t thought of. AI has been used for exactly this thing for years in chemistry. When given constraints (less lithium) and parameters (with this much capacity) it can try permutations of various designs that theoretically meet those conditions.
Yes AI is overhyped, yes it’s often exaggerated by news sources, but that doesn’t mean AI is a non-invention or something. It’s a long way off from any of the lofty goals that are often thrown around by tech ceos, but that doesn’t mean it’s useless.
We have had weather models, astronomical models, and all other kinds of computer based prediction methods that do multiple permutations that theoretically meet conditions. AI is just another step forward by doing better pattern recognition and identifying relationships with data based on design choices. All of the chemistry findings came from the system being designed to try things they would not normally test for because testing is expensive and AI can run simulated tests faster and cheaper.
My point is that saying ‘AI came up with’ is 100% inaccurate phrasing intended to trick people into thinking that AI is intelligent instead of just being a very complex tool used to do things we already do faster. It allows for trying more permutations and more pattern recognition, but is just another approach to existing computer models that have also identified things we did not expect. Computer models used to identify starts with planets, but we don’t call those intelligent because they aren’t being sold as something they are not.
Ah, I see what you’re saying. Yes the recognition for these advances should be with human programmers and engineers who are configuring the software and making the models for testing. You’re right I can definitely see why that distinction is important and the media should be making clear that the AI isn’t just turned on and magically works it all out on its own. It’s computational resources being directed towards a task, the models it works within are setup by professionals and the discoveries it finds are interpreted and made useful by those professionals.
The media is just parroting what the companies that want to sell AI are saying. They suck at reporting anything technical or scientific for sure, but they didn’t come up with this on their own.
Your first comment my first thought was how does this have any upvotes. Thats super wrong
Top notch comback with this comment, i still cant agree with the original wording, i do recognize your point and agree with yoiur sentiment. Its a tool first and foremost.
That’s not true at all. AI can in fact generate novel techniques and solutions and has already done so in biotech and electrical engineering. I don’t think you understand how AI works or what it is
I think maybe people are running into a misunderstanding between LLMs and neural nets or machine kearning in general? AI has become too big of an umbrella term. We’ve been using NNs for a while now to produce entirely new ways to go about things. They can find bugs in games that humans can’t, been used to design new wind turbine blades (even made several asymmetrical ones which humans just don’t really do), or plot out entirely new ways of locomotion when given physical bodies. Machine learning is fascinating and can produce very unique results partly because it can be set up to not have existing design biases like humans do
And the nature of computers is that they are magnitudes better than humans at brute forcing. Machine learning can brute force (depending on the technique, it can be smarter than brute forcing, being more efficient) test many many many more designs and techniques than we could manually do. Sure it’ll fail many times, but it’s just a numbers game, and it can pump those numbers. It’ll try a lot of weird and unique stuff we wouldn’t even think to try, with varying degrees of success.
What novel solutions has ai done in electrical engineering?
Name one that wasn’t just doing the thing it was told and the users being surprised. You know, the same way that people are surprised when research has results they did not expect using other approaches.
It’s a weird way of asking this. Of course it’s going to do what’s told, the alternative is that it, out of the blue, spits a battery design for no reason. If it were to somehow find a way to make batteries with less lithium in a way that never did before, isn’t that an unexpected result using other approaches?
This is not general artificial intelligence, everything we have is narrow AI, focused on solving one specific problem, for identifying birds to understand instructions between drugs.
Yeah, that would be coming up with a battery design.
That’s the point, it takes all the factors we know about and speed runs through all the possible ways it could work. Humans don’t have the time to look for every single possible way a battery could be constructed, but a ML model can just work it’s way through the issue faster and without human intervention.
Plus just like with the new group of antibiotics we just used AI to discover, it will allow truly thinking Humans to expand upon it.
Really sick of this “oh but you don’t realize AI don’t actually think! Therefore it’s all worthless!” With this smug bullshit like you think you’re bringing anything of value to the conversation.
I didn’t say it was worthless. In fact, I said the exact same things you just said in another post but with the additional detail that the name actually does matter when it is clearly misleading people into thinking it is something that it is not.
Not even close to true
Do you think AI just does things unprompted?
No one said anything about unprompted
😏
😳
Only a small subset of AI uses prompts.
Think of prompts as input
Deep learning, the most impressive type of AI, doesn’t use inputs.
What a terribly ignorant thing to say, when people make these armchair comments they’re only hurting ordinary people that can make real benefits from using the technology.
What a giant leap you have taken there. Speeding up existing processes is an extremely helpful thing for the average people, just like weather models that also did things we were already doing far faster and with more variables than people could handle without the automation.
AI will be very helpful. It will not magically solve all of our problems on its own, which is how ‘AI comes up with’ is being presented.
My favorite part is the one where you skipped over exactly what I was talking about
My favorite part was where you accused me of hurting people because I said AI does what we already do faster.
You compared AI to googling bro
I’m done with this convo lmao
By this very same logic, nobody has ever discovered anything because they’re just speeding someone else’s plans of improving or deriving from someone else’s findings
Genius.
At the core, weather models, web searches, and AI are all pattern recognition with various levels of complexity and scope. Just like a bicycle is comparable to a motorcycle because they both have two wheels even though one is powered and can go faster and for longer without wearing out the rider.
AI is not a person capable of coming up with something on its own.
I never claimed that, humans discoveries are just new permutations of observed phenomena. Every single mechanism of the universe is a permutation of the baseline functionality: physics. Therefore, if we’re shitting on permutations, you’re shitting on all of science. AI can do what we do faster. It’s just applied “knowledge” - no different than humans. In fact, that’s the whole point of neural networks, to emulate what our brains literally do right now.
Not all batteries even use lithium. So why not just go with 100% less lithium, if that’s the target metric.
SLA doesn’t get enough love. It’s still the most reliable battery type in adverse conditions, especially cold temperatures.
Just has some small issues with size, weight, and energy density.
you would know that if you read the article. they replaced part of lithium in electrolyte with sodium, so that they can use less lithium. the problem is decreased ion mobility ie less power density in real life terms.
bet
i’m gonna mostly ignore this finding because it sounds like extension of AI hype. real lab work is still absolutely critical in order to make it work
I did read it, the snippet I used is from the last part of the article…
And then there’s a hundred other factors. How many charge cycles does it get? Cold weather performance? Can it be mass produced? Does it improve safety over current cells?
It might be useful for what it leads to. Batteries get better because we explore ten different options and then one of them works out. People have gotten less excited over individual discoveries like this for mostly fair reasons. But then there’s another layer of understanding beyond that where you see it as one path of many.